A non-contact vibration detection and partial discharge fault diagnosis method for high-voltage switchgear

Abstract

To address the early concealment of partial discharge (PD) defects in high-voltage switchgear and the difficulty of external non-contact identification, this paper proposes a PD defect identification method based on laser Doppler vibrometry and vibration signal analysis. A KYN28-12 high-voltage switchgear cabinet is taken as the research object, and three typical defect types, namely protrusion discharge, surface discharge and floating discharge, are constructed. Vibration signals are collected at a unified measuring point on the enclosure by a single-point laser Doppler vibrometer. Considering the discontinuous distribution of effective discharge responses and the strong background interference in long-duration vibration records, effective discharge interval extraction and variational mode decomposition modal-selection-based mild reconstruction are adopted to enhance the vibration components related to PD. Fixed-length samples are then constructed by sliding windows, and multi-scale complexity features and impact statistical features are extracted to form a fused feature vector. A Bayesian-optimized ECOC-RBF-SVM model is used to identify the defect type. The experimental results show that the three PD defects exhibit distinguishable characteristics in the enclosure vibration response. The proposed method achieves an overall identification accuracy of 97.94% on the test set, verifying the feasibility of identifying switchgear PD defects using non-contact vibration signals and providing a reference for external condition monitoring and fault diagnosis of high-voltage switchgear.